基于强化学习的抛光机器人主动力控制研究  

Research on Active Force Control of Polishing Robot Based on Reinforcement Learning

作　　者：张活俊 江励[1] 汤健华 黄辉[1] ZHANG Huojun;JIANG Li;TANG Jianhua;HUANG Hui(Wuyi University,Jiangmen 529000,China)

机构地区：[1]五邑大学,广东江门529000

出　　处：《机械工程师》2020年第9期31-34,共4页Mechanical Engineer

基　　金：广东省省级科技计划项目“五自由度混联抛光机器人系统的开发及力控制关键技术研究”(2016A010102023)。

摘　　要：针对工业应用的抛光设备价格昂贵且对于复杂曲面工件实用性不强的问题,研发一种五自由度混联抛光机器人。首先结合主体机构和控制系统介绍抛光机器人系统,然后结合动力学仿真验证该恒力控制模型应用于抛光机器人的可行性。最后结合强化学习的深度Q网络算法求解得主动力控制策略,理论上初步证明该策略针对复杂轮廓的工件获得良好的恒力控制效果。In view of the high price of polishing equipment in industrial application and the low practicability for complex curved parts,this paper develops a five DOF hybrid polishing robot.Firstly,the polishing robot system is introduced by combining the main mechanism and control system,and then the feasibility of applying the constant force control model to the polishing robot is verified by dynamic simulation.Finally,the active force control strategy is solved by combining the deep Q Network algorithm of reinforcement learning,and it is proved that the strategy can achieve good constant force control effect for the workpiece with complex contour in theory.

关 键 词：抛光机器人 主动力控制策略 深度Q学习 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]